#141

I'd be interested in taking this up. I haven't contributed to a Rust repo before, so I'd need a bit of help with structure, but this shouldn't be too hard 🤞

For instance, there are two functions, GET_FLOAT_WORD and SET_FLOAT_WORD in the newlib implementation, and I don't know what the equivalent is for this library.

Looks like those are defined in fdlibm.h

They seem to be converting f32 to u32 and back via union type punning (don't compile that code in C++ mode!)

In Rust you'd use f32::to_bits and f32::from_bits instead.

from_bits gives back a u32, but the newlib code uses i32. I just want to make sure it's ok to use u32 instead, and if not, are the steps I need to take to guarantee that it is castable?

Converting between u32 and i32 with as will preserve the bit pattern (just not necessarily the numeric value), so feel free to use an extra as here and there if you're just trying to do a bit manipulation of a number.

So, the issue mentions moving the previous implementation to a private module. Could you point me to where I would put that? Is it in the same file, or some other part of the structure? Just an example file that has done this already would be fine, so that I can look at how they did it.

I don't see a location where a private module has been setup like that before, but I think I can describe the steps to do what @japaric was thinking:

1. Make a musl folder in src/math/
2. Move the current src/math/hypotf.rs file into that sub-folder
3. Within src/math/mod.rs place the following code after the "private re-import" portion to make sure that we keep building the old version during tests:

// Test that the old code continues to build
#[cfg(test)]
#[allow(dead_code)]
mod musl {
    mod hypotf;
}